Designing Context-Based Marketing: Product Recommendations under Time Pressure
50 Pages Posted: 23 Nov 2014 Last revised: 11 Jan 2019
Date Written: December 22, 2018
Optimal allocation of a company’s advertisement resource across different channels crucially depends on how advertisements affect consumer preference and attention respectively. Although consideration set models can be used to investigate such questions, identification of consideration set models is challenging. Low level of choice probability can be attributed either to a low level of utility or a low level of attention. In this paper, we propose a new source of identification that exploits variation in product availability, and apply the method to the field experiment data on consumer beverage purchases from vending machines. With product availability approach, we can distinguish advertisement’s effect on consumer attention and preference without relying on exclusion restrictions, which are commonly adopted in the existing literature. We find that advertisements affect both attention and preference significantly. Moreover, we find significant product-level heterogeneity in advertisement effectiveness on both preference and attention. Hence, our empirical findings deliver practical managerial implications in determining allocation of marketing resource.We study how to design product recommendations when consumers' attention and utility are influenced by time pressure---a prominent example of the context effect---and menu characteristics, such as the number of recommended products in the assortment. Using unique data on consumer purchases from vending machines on train platforms in Tokyo, we develop and estimate a structural consideration set model in which time pressure and the recommendation menu influence attention and utility. We find that time pressure reduces consumer attention but increases utility in general. Time pressure moderates the effect of recommendations for attention of both recommended and non-recommended products, and utility for recommended products. Moreover, the number of total recommendations increases consumer attention in general, but in a diminishing way. In our counterfactual simulation, we find that the revenue-maximizing number of recommendations decreases with time pressure. Optimizing the number of recommendations for each vending machine and for each time of day increases the total sales volume by 4.5% relative to the actual policy, 1.9% points more than traditional consumer-segment-based targeting.
Keywords: Consideration Set, Brand Choice, Advertising, Menu Dependence
Suggested Citation: Suggested Citation